Inversion of azimuthal electromagnetic logging while drilling data based on BiLSTM-transformer

Azimuthal electromagnetic Logging While Drilling inversion is crucial for geological evaluation in complex well conditions. However, traditional iterative inversion methods suffer from strong dependence on initial values and a tendency to converge to local optima. Furthermore, existing deep learning approaches struggle to jointly invert formation boundary distances and resistivity parameters. To address this, this study proposes an inversion method based on the BiLSTMTransformer model. This model captures local forward and backward correlation features of the logging curves through the BiLSTM network. Simultaneously, it models the spatial global correlations among formation parameters using the Transformer's multi-head attention mechanism, enabling the joint inversion of layered formation boundary distances and resistivity parameters. Experimental results demonstrate that this architecture overcomes the dilemma of deep learning inversion methods in resolving resistivity of adjacent layers and effectively enhances the inversion accuracy of formation parameters